---
title: Transcribe an audio file
type: how-to
section: ai-studio
summary: Choose Transcribe audio from the plus button next to the message box, pick a recording, review the transcript, then insert it into the thread.
applies_to:
  clients: [web]
sources:
  - evolve-front-end/app/[lng]/studio/components/Chat/audio-transcription-file.ts
  - evolve-front-end/app/[lng]/studio/components/Chat/ChatInput.tsx
  - evolve-front-end/app/i18n/locales/en/powerflow.json
verified:
  at: 2026-09-04
  env: code
  by: content-lane-ai-studio
ttl_days: 45
related: [ai-studio/use-your-voice, ai-studio/attach-files-to-a-message]
---

Select the plus button next to the message box, choose **Transcribe audio**, pick one recording, then select **Transcribe**. Review the text and select **Insert into chat** to put it in the message box.

The window states what happens to the file: **The recording stays transient. Only the selected audio and options are sent for this transcription.**

## Transcribe a recording

1. Select the plus button next to the message box, then **Transcribe audio**.
2. Select **Choose audio to transcribe** and pick one file. MP3, MP4, MPEG, MPGA, M4A, WAV, WebM, OGG and FLAC are accepted, up to **25 MiB maximum**.
3. Choose a **Transcription model**: **Best**, **Fast & economical**, **Speaker labels** if you need who said what, or **Captions, timestamps & translation**.
4. Select **Transcribe**. The status moves through **Reading audio file…** and **Transcribing…** to **Transcript ready. Review it before inserting.**
5. Correct the text under **Editable transcript**.
6. Select **Insert into chat**.

Transcribe one file at a time. Choosing several gives **Transcribe one audio file at a time.**

## Captions and translation

Choosing **Captions, timestamps & translation** adds a **Whisper task** control:

- **Transcript** — plain text in the recording's own language.
- **SRT captions** and **VTT captions** — subtitle text with timing.
- **Timestamps** — a transcript with segment times.
- **Translate to English** — English text from speech in another language.

## Help the model get names right

Open **Advanced options** and fill in what applies:

- **Languages, comma separated** for the languages spoken.
- **Keywords, comma separated** for terms it should spell correctly.
- **Names or context** for people, acronyms or specialist words.

## If it did not work

- **Choose an MP3, MP4, MPEG, MPGA, M4A, WAV, WebM, OGG, or FLAC audio file.**: convert the recording to one of those formats.
- **Audio files must be 25 MiB or smaller.**: split the recording or export it at a lower bitrate.
- **No speech was detected. Try another file or model.**: check the file plays, then try another **Transcription model**.
- **The audio could not be transcribed.**: select **Retry**.